Affiliations: [a] Department of Computer Science and Engineering, National Institute of Technology, Uttarakhand, India | [b] School of Computer and System Sciences, Jawaharlal Nehru University, New Delhi, India
Correspondence:
[*]
Corresponding author: Nitin Kumar, Department of Computer Science and Engineering, National Institute of Technology, Uttarakhand 246174, India. Tel.: +91 1346 257561; Fax: +91 1346 251095; E-mail:nitin2689@gmail.com
Abstract: Block linear discriminant analysis (LDA) is one of the
face recognition methods suggested when only single image per person is
available. However, the transformation in block LDA involves computation of
inverse of within class scatter matrix, which may not exist when within
class scatter matrix is singular. In order to overcome this, we present a
novel technique called Block LDA via QR-Decomposition. The proposed
technique does not involve the computation of within class scatter matrix.
In addition, it is also computationally efficient and scalable. The
performance of the proposed technique is compared with several other methods
in terms of average classification accuracy and training time. The proposed
technique possesses less computation complexity than several other methods
and is suitable for real time applications. Experimental results on two
publicly available datasets ORL and Yale demonstrate the efficacy of the
proposed technique.
Keywords: Complexity, extreme small sample size, illumination, linear discriminant analysis, patch
